CM retweeted
Tinkerās Dimuon is Muon with delayed diagonal conditioning, Polar Express and a rank-dependent update norm. I recovered its update rule through the API, borrowing @mrtnmās approach to StableReinforce. Hereās how š§µ
Dimuon is a new optimizer we built for LoRA fine-tuning. In our SFT tests, it often reaches lower loss in fewer steps than AdamW. We're still experimenting with Dimuon; you can Tinker with it today and share your feedback!
See our optimizers guide at tinker-docs.thinkingmachinesā¦
CM retweeted
dessa by @jack's company Block
incredibly impressive
best speech to speech model ive seen bar none
CM retweeted
new from us at @blocks
Meet Dessa, our AI model trained from scratch on nothing but speech.
We believe speech is the natural interface for technology. So we built Dessa.
And weāre inviting AI safety researchers to explore Dessa.
Learn more: block.xyz/dessa
CM retweeted
Meet Dessa, our AI model trained from scratch on nothing but speech.
We believe speech is the natural interface for technology. So we built Dessa.
And weāre inviting AI safety researchers to explore Dessa.
Learn more: block.xyz/dessa
RT @blocks: Meet Dessa, our speech model trained from scratch on human conversation.
Hear it for yourself: block.xyz/dessa httpā¦
CM retweeted
We built an 8000 sqft. high-throughput wet-lab in 30 days, to solve molecular discovery - starting with pesticides.
Most teams in drug-discovery donāt receive target-organism verification until clinical trials. Pesticides allow us to rapidly validate our model in the target organism directly. With the help of our high-throughput lab, we were able to generate compounds that killed a $13 billion pest, the Fall Armyworm.
We believe that the bitter lesson is coming for biology.
CM retweeted
It will always be the highlight of my career to have been able to see firsthand a model make progress towards the problem that got me interested in mathematics a decade ago. It was a huge honour to have been part of this team.
A uniform zero-free strip.
github.com/openai/math/blob/ā¦
CM retweeted
Astonishing how many posts like this there are about Tao, someone with no apparent arrogance who has spent his life sharing his gifts with the world.
Hey Terence
What you are experiencing right now is this weird phenomena called āhumblingā.
Most of us midwits experienced this in high school or college. If youāre lucky young adulthood. But nobody was ever smart enough to humble you. So we made a machine to ensure you get brought down to earth like the rest of us.
When you played down AI some months ago you were in the denial phase. Now you are in the anger phase. Soon you will reach acceptance like the rest of us.
CM retweeted
Last year I was laughed out of the room.
Now I wonder what Buzzard is thinking now that Claude formalized Fermat's Last Theorem!
In the coming months, it will now be possible to formalize *all* known mathematics.
CM retweeted
The Israeli military acknowledged firing on a car carrying 5-year-old Hind Rajab in Gaza after earlier denials as it opens a criminal probe into the January 2024 killing. cnn.it/3U6I6jU
CM retweeted
I donāt know who needs to hear this, but academia is not worth ruining your life over. Itās a job, that too a relatively poor paying job. Job prospects are abysmal. You lose control over where you Iive, research funds are drying up, students arenāt very interested.
CM retweeted
No opinion on math PhD but Iād still strongly recommend math major.
Most of us never reason. A situation arises, we act intuitively, then get RLāed. At best we get anxious and bounce between limited hazy alternatives.
But to prove a theorem you have to reason. You donāt know what precision or thinking really is until you have to sit and struggle to prove a theorem thatās new to you.
The crazier the world gets, the more this skill is useful. And the world is about to get very very crazy.
